Research On Forest Volume Inversion Method Based On Multi-source Remote Sensing

Forest is an important part of terrestrial ecosystem,which plays an important role in maintaining ecological security and dealing with environmental problems.Remote sensing technology has been widely used in various fields.As an indispensable part of forest remote sensing technology,it can not only obtain the data of forest resource management,but also provide a reference for the development of forest remote sensing technology,It can further reveal the ecological impact of forestry management.Forest volume is an important indicator of forest quality,which can not only measure the richness and health of forest resources,but also directly reflect the effectiveness of forest management.The advantages of spectral information in reflecting the physical and chemical characteristics of trees and the horizontal level,and the advantages of point cloud information in reflecting the vertical structure make the fusion of multi-source remote sensing data become a hot topic of forest biomass inversion.Among them,lidar is an active remote sensing technology which measures the distance between the sensor and the target through the laser emitted by the sensor.Lidar can quantitatively estimate forest parameters by actively obtaining three-dimensional coordinate information.However,due to the limitation of lidar data sources and the number of sample plots on the ground,the academic research of lidar in forestry is mostly small area,and the stand structure is relatively simple.In order to obtain better research results,the density of the selected point cloud is relatively high,which makes it difficult to apply to the production practice in the complex environment of China.In view of this,this article has carried on the research and the exploration from the following aspects:(1)Based on the airborne lidar data of medium and low density and the multi-source data of fixed sample plots on the ground,the inversion model is established according to the composition of topography and tree species to invert the forest volume.Firstly,data cleaning method is used to process the sample plot data,and then 46 height variables,10 point cloud density variables and 42 intensity variables are extracted by distributed computing,with a total of 98 statistical variables.Finally,multiple stepwise regression,random forest and other methods are used to establish the inversion model.The models were optimized and compared by cross validation and model analysis.(2)Using sentinel-2 remote sensing data and lidar data,forest vegetation classification technology based on optical multispectral and lidar three-dimensional structure information was studied.On the basis of spectral characteristic variables extracted from sentinel-2 data,the intensity and height information extracted from lidar data are fused,which has a certain guiding significance for the classification of tree species groups and effectively improves the classification accuracy.(3)Compared with the results of the national forest resources continuous inventory system based on random mechanical sampling,it can be seen that the research method can be used in the wide area forest volume measurement.In the same region and different point cloud density of Huangshan area,it can be found that the method can still maintain good results by increasing the point cloud density within the same order of magnitude and reducing the scale to districts and counties.